Ann. occup. Hyg., Vol. 46, No. 8, pp. 663–672, 2002
Published by Oxford University Press
DOI: 10.1093/annhyg/mef089
Investigation of the Aerosols Produced by a High-speed,
Hand-held Grinder Using Various Substrates
ANTHONY T. ZIMMER* and ANDREW D. MAYNARD
National Institute for Occupational Safety and Health, Division of Applied Research and Technology,
4676 Columbia Parkway, Cincinnati, OH 45226, USA
Received 2 April 2002; in final form 22 July 2002
Mechanical processes such as grinding are classically thought to form micrometer scale aerosols through abrasion and attrition. High-speed grinding has been used as the basis for testing
the hypothesis that ultrafine particles do not form a substantial component of mechanically
generated aerosols. A wide variety of grinding substrates were selected for evaluation to
represent the broad spectrum of materials available. To characterize the particle size distribution over particle sizes ranging from 4.2 nm to 20.5 µm, the aerosol-laden air collected from
an enclosed chamber was split and directed to three aerosol instruments operated in parallel.
Transmission electron microscope samples of the various grinding substrates were also
collected. The results demonstrate that ultrafine particles do have the potential to form a
significant component of a grinding aerosol for a number of substrates. It appears that the
ultrafine aerosols were formed by the following processes: (i) from within the grinding motor,
(ii) from the combustion of amenable grinding substrates and (iii) from volatilization of
amenable grinding materials at the grinding wheel/substrate interface.
Keywords: grinding; particle size distribution; ultrafine aerosols
cant as exposure metrics other than mass are considered. Indeed, much of the current discussion
surrounding the toxicity of low-solubility particles
traditionally considered to be chemically inert has
focused on particles <100 nm in diameter; commonly
termed ultrafine particles.
Very little attention has been given to the possibility of mechanical generation leading to substantial
aerosol exposure from ultrafine particles. Limited
data indicate that it is possible to generate high particle
number concentrations from mechanical processes.
McCawley et al. (2001) have shown that the numberweighted aerosol size distribution generated while
grinding beryllium ceramic is dominated by particles
<100 nm and Choe et al. (2000) have shown that dry
paint scraping and sanding can lead to substantial
particle number concentrations below 1 µm. However,
apart from these studies there is little in the literature
to indicate the likely contribution of mechanical
processes to ultrafine aerosols in the workplace.
Grinding represents a typical mechanical operation
found in many workplaces. Weight-based (e.g. gravimetric) sampling is typically conducted to characterize grinding operations (Lehmann and Frohlich,
1988; Thorpe and Brown, 1994) for comparison with
INTRODUCTION
It is generally assumed that aerosols generated during
mechanical processes such as grinding are characterized by large particles formed through attrition.
Consequently, little attention has been given to the
generated particle concentration much below 1 µm.
While it is conceivable that the mass-weighted distribution of such aerosols may be dominated by supermicrometer particles, it is by no means clear that the
same will apply for particle number- and surface
area-weighted distributions. Recent published research
has indicated that for some classes of material
(notably materials leading to aerosol particles with
low solubility) biological activity following inhalation may be more appropriately represented by
aerosol number or surface area concentration
(Oberdorster et al., 1995; Donaldson et al., 1998;
Brown et al., 2001). As the particle number or
surface area per unit mass increases with decreasing
particle size within an aerosol, the presence of submicrometer particles becomes increasingly signifi*Author to whom correspondence should be addressed.
Tel: +1-513-841-4370; fax: +1-513-841-4545;
e-mail: [email protected]
663
664
A. T. Zimmer and A. D. Maynard
various exposure criteria (ACGIH, 1990; NIOSH,
1992; OSHA, 1997). Due to the assumption that the
aerosols generated from these processes mainly result
from mechanical attrition of the substrate, prior
research to characterize the particle size distribution
has focused upon the micrometer size regime using
techniques such as cascade impaction (Lehmann and
Frohlich, 1988; Kusaka et al., 1992).
Grinding typically occurs on a wide variety of
substrates. Some substrates have no specific exposure criteria and are evaluated as a ‘particulate not
otherwise regulated’ with exposure limits ranging
from 5 to 15 mg/m3 for respirable and total exposure,
respectively (OSHA, 1997). Other substrates, such
as wood dusts and metals, can be evaluated and
compared with specific exposure criteria. Interestingly, metals such as aluminum and copper have
exposure criteria that are lower for the metal fume
than the metal dust (i.e. method of generation)
(ACGIH, 1990; OSHA, 1997).
In this study, high-speed grinding was used to test
the hypothesis that ultrafine particles do not form a
substantial component of mechanically generated
aerosols. Aerosols from the grinder were produced in
a controlled environment, using a variety of grinding
substrates, and the grinding aerosol was characterized
over a particle size range from 4.22 nm to 20.5 µm.
MATERIALS AND METHODS
HEPA-filtered, particle-free air was pulled upward
through a stainless steel chamber (0.914 m long,
0.813 m wide, 1.19 m tall) using a commercially available, industrial air cleaner (model 73-800G; ACE
Corp., St Louis, MO) (Fig. 1). Prior to measurements,
the air cleaner was turned on and the HEPA-filtered
air within the chamber was monitored using a
condensation particle counter (CPC) (model 3022A;
TSI Inc., Shoreview, MN). When the particle number
concentration was reduced to a value of ∼50
particles/cm3, the air cleaner was turned off to
prevent flow through the chamber. A small, tubeaxial
fan (model 8500C; Pamotor, Burlingame, CA),
located within the chamber, was used to create
‘stirred’ conditions within the chamber (CPC measurements validated that the fan did not represent a
source of aerosols within the chamber).
Grinding was accomplished using a Multipro™
(model 395; Dremel, Racine, WI), a variable speed
tool with rotational speeds that can be varied from
5000 to 30000 r.p.m. A cylindrical grinding wheel,
composed of sintered aluminum oxide, was selected
for these experiments (diameter 1.6 cm, length 1.0 cm).
To prevent cross-contamination, new grinding wheels
were used for each substrate tested. The rotational
speed of the grinding wheel was determined using a
tachometer (Dynapar model HT50; Danaher Controls,
Gurnee, IL). Grinding was accomplished such that
the cylindrical wheel was placed normal to the
substrate with a constant applied force of 3.96 N. A
wide variety of grinding substrates were selected for
evaluation to represent the broad spectrum of materials available. The substrates selected included both
relatively homogeneous and heterogeneous materials
Fig. 1. Schematic of the grinding generation and aerosol sampling apparatus. Note that the shut-off valve closed when generating
and sampling grinding aerosols.
Aerosols produced by a high-speed, hand-held grinder
and included steel, aluminum, polytetrafluoroethylene (PTFE), hardwood, granite and clay ceramic.
A number of aerosol instruments were used to
evaluate the aerosol behavior and the resultant
particle size distribution during a grinding operation.
Using stainless steel and flexible graphite tubing to
attenuate charging effects within the sample line, the
grinding aerosols were collected from the stainless
steel chamber from a position located directly above
the grinding operation (height 15 cm). To ensure that
there was little temporal variation over the time
scales needed to collect aerosol samples, a CPC was
used to characterize the temporal variation of the total
number concentration during a grinding sample run.
To characterize the particle size distribution, the
aerosol-laden air from the chamber was split and
directed to three aerosol instruments operated in
parallel. Small, nanometer scale particles [4.22 nm <
dp (particle diameter) < 100 nm] were characterized
using a scanning mobility particle sizer (SMPS)
configured with a nano differential mobility analyzer
(DMA) (electrostatic classifier model 3080 using a
DMA model 3085 and a condensation particle counter
model 3022A; TSI Inc., St Paul, MN). Larger,
nanometer scale particles (60.4 nm < dp < 777 nm)
were characterized using a SMPS configured with a
long DMA (DMA model 3934 and a condensation
particle counter model 3022A; TSI Inc.). Larger,
primarily micrometer scale particles (523 nm < dp <
20.5 µm) were characterized using an aerodynamic
particle sizer (APS) (model 3320; TSI Inc.). Rogak et
al. (1993) have shown particle mobility diameter to
agree well with equivalent-sphere projected area
diameter for fractal-like particles <1 µm. The assumption was therefore made that the data from each SMPS
could be interpreted in terms of particle equivalentsphere projected area diameter. Particles large enough
to be sampled by the APS were assumed to arise
predominantly through attrition and therefore to have
a compact morphology. APS aerosol size distributions were therefore transformed to particle number
concentration versus equivalent-sphere projected area
diameter assuming spherical particles with the same
density as the bulk substrate material. Furthermore,
the APS data were corrected for sampling train losses
between the chamber and the instrument inlet and
losses within the APS nozzle (Kinney and Pui, 1995).
Calculations indicated sampling train losses to each
SMPS to be negligible. The total sample time for
each SMPS measurement was 230 s (up-scan time
200 s, down-scan time 30 s) while the total sample
time for each APS measurement was 200 s.
Transmission electron microscope (TEM) samples
of the aerosol generated from each substrate type
were also collected using a point-to-plane electrostatic precipitator (Cheng et al., 1981).
Triplicate, randomized samples were obtained for
each grinding substrate. During a typical experi-
665
mental sample run, the grinding substrate was
attached to a holder. The grinding wheel was placed
on the substrate and the grinding speed was adjusted
to 20000 r.p.m. using a rheostat (model 3PN751;
Matheson Scientific) and a tachometer. The chamber
was HEPA filtered to remove aerosols within the
chamber and the ventilation system was turned off.
After ∼30 s, grinding was conducted on the substrate
for 10 s, ensuring that the aerosol number concentration was sufficiently low to suppress coagulation
dynamics (typical number concentrations were <106
particles/cm3). Simultaneous aerosol sampling (i.e.
SMPS/NDMA, SMPS/LDMA and APS) was started
60 s after the grinding burst. Sampling was started at
this point, based upon CPC measurements suggesting
that the aerosol within the chamber was well mixed
and that the particle number concentration did not
vary appreciably over the necessary time scales.
To ensure that the aerosols measured representative products of grinding, background aerosol
measurements were also taken with the grinding tool
freely spinning within the chamber at 20000 r.p.m.
with an unused sintered aluminum oxide grinding
wheel. Given the unlikely probability that aerosols
were created from the grinding wheel, this measurement represented the aerosols produced primarily by
the grinding tool motor. This result was subtracted
from measured size distributions for each grinding
substrate.
RESULTS
The aerosol produced solely by the grinding tool
and the background corrected measurements for each
of the substrates are presented in Fig. 2a–g, with electron micrographs of representative particles shown in
Fig. 3a–e. In many cases the measured size distributions exhibited multi-modal behavior. The measurements obtained using each aerosol instrument are
represented by different symbols (i.e. solid circles
for the SMPS/NDMA configuration, hollow circles
for the SMPS/LDMA configuration and solid triangles
for the APS). Each point in the figures represents the
arithmetic mean of the triplicate random samples
while the bars estimate the standard error associated
with each particle size class.
To quantitatively present the particle size statistics
for the combined aerosol instruments, a log-normal
particle size distribution was assumed and fitted to
the ultrafine and coarse modes in the experimental
data (Fig. 2a–g and Table 1). In presenting number
statistics for the ultrafine mode, the data obtained by
the SMPS/NDMA were used for fitting the left-hand
side of this mode as this instrument was optimized to
sample aerosols in the particle size range 10–100 nm.
In presenting number statistics for the coarse mode,
the data obtained by the APS were fitted using the
right-hand side of this mode. As can be seen, the
666
A. T. Zimmer and A. D. Maynard
Fig. 2. Variation in particle number concentration [dN/d log(dp)] as a function of equivalent-sphere projected surface area diameter
for (a) the Dremel™ tool without a grinding substrate, (b) a granite substrate, (c) a clay ceramic substrate, (d) a steel substrate, (e)
an aluminum substrate, (f) a PTFE substrate and (g) a hardwood substrate using the SMPS/NDMA (solid circles), SMPS/LDMA
(hollow circles) and APS (solid triangles). Note that (i) the solid line for each experimental condition represents a log-normal fit of
the experimental data, (ii) the error bars represent the standard error associated with three randomized replicate samples and (iii)
the experimental data for all substrates have been background corrected for the Dremel™ tool aerosols.
SMPS/LDMA tended to sample particle sizes corresponding to aerosols located between the ultrafine and
coarse modes. Sampling errors associated with this
instrument tended to be larger than those associated
with either of the other instruments above 200 nm,
reflecting the low particle detection rate within the
instrument in this regime.
The particle size distribution produced by the
Dremel™ tool alone (Fig. 2a) appears to be slightly
bimodal. In the overlap regions there was relatively
good agreement among the aerosol instruments.
Grinding upon the granite substrate gave a distri-
bution that was relatively bimodal (Fig. 2b). TEM
analysis indicated a large number of particles >1 µm
in diameter, characterized by compact non-spherical
particles with sharp edges indicative of generation
through abrasive removal (Fig. 3a). The micrograph
also revealed evidence of large numbers of nanometer size particles on the substrate, although it is not
clear whether these originated from the granite
substrate or the grinding tool itself.
Grinding of clay ceramic led to a distinct bimodal
particle size distribution (Fig. 2c). Some discrepancy
was observed in the overlap region between the
Aerosols produced by a high-speed, hand-held grinder
667
Fig. 2. Continued.
APS and the SMPS/LDMA. Both instruments showed
decreased detection efficiency for particles in the
region of 500 nm and it is likely that this was the
source of the observed mismatch. Interpolation
between data points below 400 nm and above 800 nm
gave a smooth distribution. TEM images were characterized by large numbers of compact particles
>1 µm, with some evidence for abraded particles
<<1 µm (Fig. 3b).
The aerosol produced when grinding on a steel
substrate is shown in Fig. 2d. When compared with
the aerosols produced solely by the grinding tool, the
total number concentration for this substrate was over
one order of magnitude higher, although qualitatively
similar to that produced using only the Dremel™ tool
(Fig. 2a). Grinding of the steel substrate led to a
distinct difference between the results obtained in the
overlap region between the SMPS/LDMA and the
APS. The high density particles produced from the
steel substrate (ρ ≈ 7.9 kg/m3), together with uncertainty over their morphology, are probably responsible
for the poor transformation between aerodynamic
diameter and equivalent-sphere projected area diameter with the APS data. Since the SMPS tends to
overestimate particle counts at high diameters when
particle counts at lower diameters are significantly
higher (an artifact associated with particle residence
time within the system), it is conceivable that, in the
668
A. T. Zimmer and A. D. Maynard
Fig. 2. Continued.
absence of such an artifact, better agreement between
the SMPS and APS would be seen. In addition, the
APS response begins to drop off markedly below
∼0.7 µm aerodynamic diameter, corresponding to a
physical diameter of ∼2 µm with a particle density of
∼7.9 kg/m3. Thus it is likely that the observed
mismatch is attributable to instrument response and
conversion errors between particle diameters. TEM
images were characterized by open chain-like agglomerates of very fine primary particles (Fig. 3c), indicating nucleation followed by growth through
coagulation. These images also indicate that many of
the particles sampled by the APS may not have been
compact (as was assumed).
A similar aerosol size distribution was measured
for the aluminum substrate, although no comparable
discrepancy between the SMPS/LDMA and APS was
seen (Fig. 2e). The behavior of this substrate was
qualitatively similar to that produced by the grinding
tool itself (Fig. 2a). However, the total number concentration when grinding upon an aluminum substrate
was over 18 times higher. TEM images showed large
compact particles (Fig. 3d) and many more nanometer size particles. Once again, it was not possible
to distinguish a source and generation mechanism for
the particles by morphology alone.
The aerosol that resulted from grinding a PTFE
substrate is shown in Fig. 2f. Large numbers of single
Aerosols produced by a high-speed, hand-held grinder
669
Fig. 2. Continued.
particles around 10–20 nm in diameter were seen by
TEM (Fig. 3e). Although it was not possible to positively identify these particles, they probably arose
from grinding of PTFE because no comparable
particles were seen in any of the other samples.
Finally, Fig. 2g shows the measured aerosol size
distribution for grinding of oak hardwood. While still
bimodal, the lower mode occurs at a much higher
diameter than for previous substrates (Table 1). When
comparing the instrument response in the overlap
region between the SMPS/LDMA and the APS, a
distinct difference was noted between the measured
results. As in the case of steel, it is likely that uncertainty over the density of the wood-related aerosol
and the particle morphology were responsible for a
poor transformation between particle aerodynamic
diameter and equivalent-sphere projected area diameter. The aerosol collected for TEM was, for the
most part, too beam sensitive to allow imaging. The
lack of non-organic carbonaceous particles indicated
that most of the ultrafine aerosol was associated
with incomplete combustion and probably contained
a large number of volatile components that evaporated in the TEM.
DISCUSSION
The aerosol number concentrations generated by
the Dremel™ tool on the substrates tested varied
greatly (Table 1). The particle size distributions
tended to be bimodal, with many substrates showing
distinct ultrafine and coarse modes. In the overlap
region between the SMPS/NDMA and SMPS/LDMA,
there was good agreement between the experimental
results, demonstrating the usefulness of using the
two SMPSs in parallel to measure the particle sizes
from 5 to 700 nm. In the overlap region between the
SMPD/LDMA and APS, the visual agreement of the
size distributions obtained with the two instruments
was reasonably good following transformation of
aerodynamic diameter to an estimated equivalentsphere projected area diameter, and only varied
appreciably for substrates of very high or low density.
When comparing the particle size distributions for
the various substrates, it appears that the coarse modes
varied markedly (Fig. 4). Grinding upon a surface
such as ceramic or granite produced an aerosol with
appreciable number concentrations and a distinct
mode, while grinding upon surfaces such as aluminum
produced low particle number concentrations and an
indistinct mode. Additionally, it appears that the
count median particle diameters varied markedly from
substrate to substrate, with no discernable pattern.
When viewing all of the substrates together for
particle diameters corresponding to the ultrafine
mode, it can be seen that, with the exception of wood,
the size distributions generated by grinding were
qualitatively similar to that produced using the
grinding tool alone (Fig. 5). Examining the electron
micrographs from each substrate gave little indication of whether particles in the ultrafine mode were
primarily associated with the substrate or the carbon
brush motor within the grinding tool. Grinding of
PTFE was the main exception; this produced large
numbers of nanometer diameter, nearly spherical
particles that were almost certainly PTFE. Also,
micrographs from the steel substrate showed the
existence of nanometer size primary particles within
large open agglomerates. However, for the aluminum,
ceramic and granite substrates the source(s) of the
670
A. T. Zimmer and A. D. Maynard
Fig. 3. Transmission electron microscope micrographs of the various grinding substrates including: (a) granite, (b) clay ceramic,
(c) steel, (d) aluminum and (e) PTFE. Note that the hardwood TEM sample did not image well, indicating aerosols composed of
volatile organic hydrocarbons.
Aerosols produced by a high-speed, hand-held grinder
671
Table 1. Size distribution statistics weighted by particle number for the Dremel™ tool operating by itself and when grinding upon
various substrates
Experimental condition
Ultrafine mode
Ntot (cm–3)
Coarse mode
dCMD (nm)
σg
Ntot (cm–3)
dCMD (nm)
σg
Dremel™ without substrate
6.5 ×
103
9
2.0
3.0 × 100
600
1.5
Granite
1.0 × 104
10
1.7
1.9 × 101
800
1.7
Ceramic
5.0 × 104
10
1.7
4.0 × 102
700
1.9
Steel
8.0 × 104
10
1.7
6.0 × 100
850
1.5
Aluminum
1.2 × 105
11
1.8
5.0 × 100
600
1.8
PTFE
8.0 × 104
10
1.7
8.5 × 10–1
2000
1.5
Hardwood
8.0 × 105
35
1.7
1.2 × 101
3800
1.5
Particle statistics represent average values for three replicate sample runs and the reported statistics for each substrate have been
background corrected.
Fig. 4. Composite of the coarse mode experimental results for the Dremel™ tool and various grinding substrates. The information
is extracted from Fig. 2a–g.
nanometer size particles could not be discerned. In
the case of PTFE and steel, the ultrafine particle
morphology indicates formation through nucleation,
suggesting that temperatures were sufficiently high
at the grinding interface to lead to vaporization or
combustion of the substrate. Grinding on wood
resulted in an ultrafine aerosol with a markedly larger
median diameter and substantially higher number
concentration. It showed evidence of charring at the
grinding interface, supporting the hypothesis that
frictional heating during grinding can lead to the
formation of ultrafine particles.
Clearly, a better indicator of the source of an
ultrafine particle would be obtained through single
particle elemental analysis. Energy dispersive X-ray
analysis (EDX) is widely used in transmission electron microscopy to identify elemental composition.
However, when applied to nanometer sized particles
EDX is limited by long sample collection times and
the sensitivity of the sample to beam damage and
contamination. In this case, it was not possible to
obtain meaningful information from the samples
using EDX. For a better understanding of the source
of the ultrafine mode arising from grinding on these
substrates, the use of alternative sample collection
methods or analytical methods such as electron
energy loss spectroscopy (EELS) should be considered for future research (Brydson, 2000; Maynard,
2000).
Although these data raise some ambiguity over the
source of the ultrafine particles, they also raise the
question of whether the process leading to mechanical aerosol generation should be considered a source
of particles in its own right. A number of publications
have indicated that for low solubility, chemically
‘inert’ materials the composition of ultrafine particles
is of secondary importance (Oberdorster et al., 1995;
Donaldson et al., 1998; Brown et al., 2001). If this is
672
A. T. Zimmer and A. D. Maynard
Fig. 5. Composite of the ultrafine mode experimental results for the Dremel™ tool and various grinding substrates. The
information is extracted from Fig. 2a–g.
the case, there is little to be gained in differentiating
between particles generated by the tool and those
from the substrate. However, in the case of highly
toxic particles such as silica, it would be important to
differentiate between the aerosols generated by these
two sources. Clearly, there is further research to be
carried out in this area.
CONCLUSIONS
In this preliminary study of particle formation upon
high-speed grinding we have shown that ultrafine
particles are generated by a variety of substrates.
There is evidence that grinding of some substrates
produced ultrafine particles through vaporization or
combustion of the substrate material. In the absence
of a substrate, there was also evidence that the
grinding tool led to the formation of ultrafine particles,
albeit at a much lower concentration. Further research
into the chemical nature of individual particles is
required to elucidate the relative source contributions
during grinding operations.
REFERENCES
ACGIH. (1990) Threshold limit values for chemical substances
and physical agents. Cincinnati, OH: American Conference
of Governmental Industrial Hygienists.
Brown DM, Wilson MR, MacNee W, Stone V, Donaldson K.
(2001) Size-dependent proinflammatory effects of ultrafine
polystyrene particles: a role for surface area and oxidative
stress in the enhanced activity of ultrafines. Toxicol Appl
Pharmacol; 175: 191–9.
Brydson R. (2000) A brief review of quantitative aspects of
electron energy loss spectroscopy and imaging. Mater Sci
Technol; 16: 1187–98.
Cheng Y-S, Yeh H-C, Kanapilly GM. (1981) Collection efficiencies of a point-to-plane electrostatic precipitator. Am Ind
Hyg Assoc J; 42: 559–632.
Choe KT, Trunov M, Grinshpun SA et al. (2000) Particle
settling after lead-based paint abatement work and clearance
waiting period. Am Ind Hyg Assoc J; 61: 798–807.
Donaldson K, Li XY, MacNee W. (1998) Ultrafine (nanometer)
particle mediated lung injury. J Aerosol Sci; 29: 553–60.
Kinney PD, Pui DYH. (1995) Inlet efficiency study for the
TSI Aerodynamic Particle Sizer. Part Part Syst Char; 12:
188–93.
Kusaka Y, Kumagai S, Kyono H, Kohyama N, Shirakawa T.
(1992) Determination of exposure of cobalt and nickel in the
atmosphere in the hard metal industry. Ann Occup Hyg; 36:
597–607.
Lehmann E, Frohlich N. (1988) Particle size distribution of
wood dust at the workplace. J Aerosol Sci; 19: 1433–5.
Maynard AD. (2000) Overview of methods for analysing
ultrafine particles. Phil Trans R Soc A; 358: 2593–609.
McCawley MA, Kent MS, Berakis MT. (2001) Ultrafine beryllium number concentration as a possible metric for chronic
beryllium disease risk. Appl Occup Environ Hyg; 16: 631–8.
NIOSH. (1992) Recommendations for occupational safety and
health: compendium of policy documents and statements.
Cincinnati, OH: National Institute for Occupational Safety
and Health.
Oberdorster G, Gelein RM, Ferin J, Weiss B. (1995) Association of particlulate air pollution and acute mortality:
involvement of ultrafine particles? Inhal Toxicol; 7: 111–24.
OSHA. (1997) OSHA Table Z-1-A. Washington, DC: Occupational Safety and Health Administration.
Rogak SN, Flagan RC, Nguyen HV. (1993) The mobility and
structure of aerosol agglomerates. Aerosol Sci Technol; 18:
25–48.
Thorpe A, Brown RC. (1994) Measurements of the effectiveness of dust extraction systems of hand sanders used on
wood. Ann Occup Hyg; 38: 279–302.